# MCPify your n8n workflows - the easy way! Source: https://www.youtube.com/watch?v=XqBbypb-Zdk Recap page: https://rapidrecap.app/video/XqBbypb-Zdk Generated: 2025-11-27T19:05:33.263+00:00 --- ## Quick Overview The video demonstrates how to use the n8n-mcp tool to execute n8n workflows remotely from any connected n8n instance, such as a local Cloud Code instance, by searching for workflows, retrieving their details, and executing them with specific inputs, successfully running a customer support workflow via a webhook trigger as a proof of concept. **Key Points:** - The presenter demonstrates executing an n8n workflow remotely using the `n8n-mcp` tool via the command line interface (0:05). - The initial test involved executing the 'Customer Support' workflow using the input 'HI', which successfully returned a greeting response (0:09, 0:16). - The workflow execution process involves three steps using `n8n-mcp`: `Search Workflows (MCP)`, `Get Workflow Details (MCP)`, and `Execute Workflow (MCP)` (0:38). - The presenter shows the output of `mcp` command (2:51), listing connected MCP servers, including the configured 'n8n-mcp' server which has read-only access for searching/details and destructive access for execution (3:06). - The customer support workflow utilizes a complex RAG pipeline involving a `Customer Support Agent` that delegates tasks to planning, transcription, and verification agents (2:01). - The final test executes the workflow using two inputs from a local CSV dataset ('final-eval-50 - Sheet1.csv') in parallel, demonstrating batch execution capability (7:42, 8:24). - The successful execution of the test cases validates the ability to leverage the n8n-mcp functionality for remote workflow execution and evaluation directly from a local development environment (9:34). ![Screenshot at 0:09: The initial command interaction shows the presenter using 'n8n-mcp' to execute the customer support workflow with the input 'HI', immediately followed by the successful execution response and workflow details.](https://ss.rapidrecap.app/screens/XqBbypb-Zdk/00-00-09.png) **Context:** The video serves as a tutorial demonstrating the capabilities of a tool called 'MCPify' or `n8n-mcp`, which extends the functionality of n8n workflows by allowing them to be controlled and executed remotely from a separate, connected n8n instance (the MCP server). The presenter uses this integration to run a complex, existing 'Customer Support' workflow that utilizes multiple specialized AI agents for RAG (Retrieval-Augmented Generation) tasks, showcasing the utility of this remote execution capability for testing and evaluation. ## Detailed Analysis The video provides a tutorial on using the `n8n-mcp` command-line tool to manage and execute n8n workflows hosted on a remote n8n instance (the MCP server) from a local environment, referred to as 'Cloud Code' in the demonstration (0:05). The presenter begins by showing a successful execution of a 'Customer Support' workflow triggered by a simple 'HI' input, which correctly classified the input as a greeting (not a support question) and returned a friendly response (0:16). The core process of remote execution is broken down into three tool commands: searching for the workflow using a query, getting its specific details (including its webhook trigger structure), and finally executing it with the necessary inputs (0:38). The presenter confirms the 'n8n-mcp' server connection is active and lists the available tools: Search Workflows (read-only), Execute Workflow (destructive/open-world), and Get Workflow Details (read-only) (2:51). The workflow itself is a complex RAG pipeline featuring a central 'Customer Support Agent' that orchestrates planning, transcription retrieval from Pinecone, and verification agents (2:01). The presenter then demonstrates a more advanced use case: running two test inputs simultaneously from a local CSV file ('final-eval-50 - Sheet1.csv') against the remote workflow (7:42). The execution shows that both requests run in parallel, and the final output, viewed in the 'Respond to Webhook' node details, confirms the complex agents executed correctly, retrieving relevant course transcriptions to answer the test queries (4:08). The presenter concludes by highlighting the power of this integration, which allows for easy automation of testing and optimization of remote workflows without being restricted by the API's limitations, suggesting this functionality is useful for running large-scale evaluations (8:53). ### MCP Connection and Tools - Listing connected MCP servers ('n8n-mcp' is connected) - Available tools are Search Workflows (read-only), Execute Workflow (destructive), and Get Workflow Details (read-only) - The tool enables remote execution of workflows - The presenter uses 'n8n-mcp' to initiate the remote execution (2:51, 3:06). ### Initial Workflow Test - Executing 'Customer Support' workflow with input 'HI' - Workflow correctly identifies input as 'UNRELATED' greeting, not a support question - Workflow returns a friendly greeting response (0:09, 0:16). ### Workflow Structure Deep Dive - The 'Customer Support v3 - Final' workflow uses a webhook trigger expecting a POST request with 'user_query' in the body - The workflow involves an LLM Router, Edit Fields, and a central 'Customer Support Agent' orchestrating sub-agents like Planning Agent and Transcriptions Retriever (2:01). ### Batch Evaluation using Local Data - The presenter reads the first two inputs from 'final-eval-50 - Sheet1.csv' - Two parallel `n8n-mcp - Execute Workflow` commands are issued using these inputs (7:42, 8:24) - Both executions run concurrently, demonstrating batch testing capability (8:26). ### Evaluation Results and Analysis - The output shows both tests executed successfully, providing detailed responses and execution logs for both test cases (9:34) - Test 1 addressed a Flowise installation issue - Test 2 correctly distinguished between AI Agents and LLMs (9:52). ### Conclusion on Utility - The ability to execute and evaluate remote workflows directly from the local environment using built-in tools is powerful for development and testing, allowing for iteration beyond direct API calls (8:53, 10:13). ![Screenshot at 0:00: The presenter is in a terminal environment, initiating an interaction with the n8n-mcp server to search for and execute a customer support workflow using the input 'HI'.](https://ss.rapidrecap.app/screens/XqBbypb-Zdk/00-00-00.png) ![Screenshot at 0:18: The workflow execution graph is displayed, showing the 'Customer Support v3 - Final' structure, which includes a webhook trigger, an LLM Router, and the central 'Customer Support Agent' coordinating sub-agents.](https://ss.rapidrecap.app/screens/XqBbypb-Zdk/00-00-18.png) ![Screenshot at 0:49: The presenter navigates to 'Settings' and then 'MCP Access' in the user interface to show the configuration settings for enabling MCP access and connecting to the server.](https://ss.rapidrecap.app/screens/XqBbypb-Zdk/00-00-49.png) ![Screenshot at 7:44: The screen displays the content of the 'final-eval-50 - Sheet1.csv' file, which contains structured input/output pairs used for batch testing the workflow.](https://ss.rapidrecap.app/screens/XqBbypb-Zdk/00-07-44.png) ![Screenshot at 9:34: The terminal output shows the results of running two test cases in parallel using the data from the CSV file, confirming both workflows executed correctly and provided expected outputs.](https://ss.rapidrecap.app/screens/XqBbypb-Zdk/00-09-34.png)